Federico Beffa <be...@ieee.org> writes:
> On Thu, Jul 9, 2015 at 1:19 AM, Mark H Weaver <m...@netris.org> wrote:
>> emacs-mmm-mode is failing to build on hydra, because the downloaded
>> tarball doesn't match the expected hash:
>>
>> http://hydra.gnu.org/build/563640/nixlog/1/tail-reload
>>
>> --8<---------------cut here---------------start------------->8---
>> starting download of
>> `/gnu/store/gdwf9rsqvvqqb77qkgpp94w8k8z1i6wf-mmm-mode-0.5.4.tar' from
>> `http://stable.melpa.org/packages/mmm-mode-0.5.4.tar'...
>>
>> http://stable.melpa.org/.../mmm-mode-0.5.4.tar 0.0% of 300.0 KiB (0.
>> KiB/s)
>> http://stable.melpa.org/.../mmm-mode-0.5.4.tar 21.3% of 300.0 KiB (405.
>> KiB/s)
>> http://stable.melpa.org/.../mmm-mode-0.5.4.tar 42.7% of 300.0 KiB (372.
>> KiB/s)
>> http://stable.melpa.org/.../mmm-mode-0.5.4.tar 64.0% of 300.0 KiB (439.
>> KiB/s)
>> http://stable.melpa.org/.../mmm-mode-0.5.4.tar 85.3% of 300.0 KiB (411.
>> KiB/s)
>> http://stable.melpa.org/.../mmm-mode-0.5.4.tar 100.0% of 300.0 KiB (419.
>> KiB/s)
>> output path
>> `/gnu/store/gdwf9rsqvvqqb77qkgpp94w8k8z1i6wf-mmm-mode-0.5.4.tar'
>> should have sha256 hash
>> `1llkzb6d978ym3zv3yfzwj0w5zzmmj3ksrm5swrx1papxcnqnkb9', instead has
>> `1kjc41nlsf7qxmmy9mrzk6myinjvc550zl5ia0ivvdz945x39yay'
>> --8<---------------cut here---------------end--------------->8---
>>
>> What went wrong here?
>
> That's not a good sign. I still have in the store the tar file with
> the hash indicated in the package and it builds fine. However, if I
> 'guix download ...' the file again I get a hash which is different
> from the one in the package and also from the one in the hydra log.
>
> I can imagine that melpa is building the tar file on the fly. Is guix
> taking the time stamp as an input to determine the hash?
Guix is taking a hash of the complete contents of the downloaded file,
i.e. all of the bytes inside the file. That, of course, includes
timestamps from inside the tar file, but not any metadata of the file
itself.
Mark
